EMC Offers Hardware-Based HA Alternative

July 24, 2007 Alex Woodie

While EMC is best known in the System i world as the platform’s only other provider of external disk besides IBM, a surprising number of EMC’s System i customers also utilize the company’s high availability offerings. What’s more, customers can use EMC’s HA and DR offerings, called TimeFinder and SRDF, not only to augment their System i servers, but to protect their Unix, Linux, NetWare, and Windows servers, thereby reducing complexity while boosting resiliency.

There was a time when only the biggest shops in the land could afford to even think about investing in a storage area network (SAN), such as EMC’s Symmetrix arrays. With price tags that start in the mid six digits and range upward into the tens of millions of dollars, the move to a centralized storage architecture is not to be taken lightly.

But times have changed. Today, mid size organizations are facing many of the same problems that the largest shops dealt with five and 10 years ago, such as the proliferation of server operating systems and an ever-increasing volume of data to store. Some mid size businesses are also affected by new industry mandates requiring more precise management of data, which further pushes shops toward central storage.

And of course, external disk brings capabilities for high availability (HA) replication and disaster recovery (DR), which are things that companies of all sizes are dealing with. After all, when an organization has consolidated all of its data assets for multiple, disparate applications and server platforms, it behooves them to avoid having a single point of failure. EMC has been offering hardware-based HA and DR solutions for its Symmetrix customers for many years.

EMC’s HA and DR solutions for Symmetrix are called TimeFinder and Symmetrix Remote Data Facility (SRDF). TimeFinder is EMC’s local replication solution, and as such it is used to minimize server downtime during backups, as well as for copying libraries for testing purposes. When a customer needs to get its Symmetrix data offsite as part of its DR strategy, it employs SRDF, which enables data banked with TimeFinder to be replicated from one Symmetrix array to another.

EMC offers both synchronous or asynchronous replication with SRDF. Synchronous has the advantage of ensuring that disk writes occur on both the local and remote Symmetrix array at the same time, which eliminates data exposure, but it only works over a short distance, and it also brings a performance hit. Asynchronous replication carries a slight risk of data exposure (because data is first written to the local disk then to the remote disk), but it works over an unlimited distance, and it doesn’t affect performance.

Customers can mix and match synchronous and asynchronous replication with their SRDF deployments, and whichever solution they chose, EMC offers Symmetrix an array of deployment options, including: bi-directional; source/target swap; one-to-many; many-to-one; and many-to-many replication.

According to Rick Aguiar, the global practice manager for the System i in EMC’s Global Services division, more than 60 percent of EMC’s System i Symmetrix customers are using TimeFinder or SRDF to boost their availability. Aguiar says EMC sells about 100 to 120 new Symmetrix arrays each year to System i-using customers (both new customers and existing customers). While Aguiar chose not to disclose the size of the System i Symmetrix user base, a rough guess puts it around 500 to 700 System i shops, which means about 300 to 400 System i shops around the world are using EMC’s high availability solutions. That’s not going to challenge Vision Solutions for supremacy in the i5/OS HA market, but it’s a fairly decent user base for an offering that EMC rarely talks about or publicizes.

That may be a higher number than one would initially expect, but keep in mind that these aren’t “pure-play” System i shops either, according to Aguiar. “These customers are looking to for high availability solutions for their Windows NT or Oracle environments, and at the same time they have an iSeries for some applications, and now they’re looking for a solution for all of those platforms,” he says. “More customers are attracted to hardware-based replication because these are open systems, so everybody in the data center uses the same concepts. … All the storage is on the Symmetrix and we replicate it.”

Aguiar says there are several advantages to using a hardware-based replication solution such as EMC’s TimeFinder and SRDF compared to software-based high availability solutions, which are sometimes called logical replication solutions, and which Aguiar refers to as journal solutions.

In particular, Aguiar has seen organizations with large System i deployments struggle to make software-based HA work for them. Getting a good “synch point” can be very difficult to achieve in large deployments, he says, and it’s compounded by the lag time of restarting System i servers–he has seen it take upward of 40 minutes to restart a System i Model 595. Five- to six-hour role swaps are not unheard of. “[The synch point] is a moving target, so they can’t consistently meet that requirement,” he says. “The recovery time for hardware-based HA or DR is much more consistent.”

But Aguiar does not pretend that EMC’s hardware-based HA will be a panacea for all of a customer’s availability problems. In many cases, a hybrid approach works best, he says. “We don’t suggest we come in and discard what you have. I look at what’s best for the customer,” Aguiar says. “For many customers, it’s the remote replication for DR that’s really a challenge. In many cases, we leave the journal solution in place for HA and do hardware replication for DR.”

Of course, EMC isn’t the only game in town when it comes to hardware-based replication. IBM also offers hardware-based replication to its System i, System p, and System z customers, through the FlashCopy, MetroMirror, and GlobalMirror solutions that work with its DS8000 series of SAN arrays (see Don’t Overlook Hardware-Based High Availability Alternatives).

But Aguiar is quick to point out differences between the HA and DR offerings of EMC and IBM. First of all, EMC’s offerings have been on the market several years longer than IBM’s, he says. And he adds, it wasn’t until members of the Large User Group–a group composed of the world’s largest System i customers that has considerable sway with IBM–started using EMC’s storage and replication offerings that IBM figured it had better bone up on external storage.

That extra time on the market has allowed EMC to build a bigger ecosystem around SRDF and TimeFinder, Aguiar says. “They are much more mature, have been out there much longer, and are accepted easier by open system players,” he says. “EMC has the advantage with its E-Lab certification. We can test and qualify operating systems . . . very quickly, make it a plug and play. So when looking at large customers, once you factor in the NT and Intel architecture, EMC as the advantage.”

EMC’s Symmetrix arrays also offer technological advantages over IBM’s DS8000 SAN arrays, Aguiar says. While both arrays utilize IBM’s 64-bit Power processors, they have significant architectural differences, he says. Namely, Aguiar points to the Direct Matrix Architecture introduced with the latest DMX series of Symmetrix arrays, which provides throughput up to 64 GB per second. DMX arrays function fine up to the 90 percent rate, which is considerably more than IBM can get out of its DS8300 series, he says. “I think they’re very, very different from a design perspective,” he says.